Best Of Product Roundup - September 12, 2026

The best expert-curated product content from the best product thinkers from around the Internet delivered twice a month.

Every two weeks, we dive into the vast ocean of product content online, both free and paid. We sift through it all to bring you the crème de la crème. Our round-up includes carefully curated links and concise excerpts from top content.

This week’s highlights feature:

• Top 3 Long-Form Articles/Podcasts

• Top 3 Tweets

• Top 4 LinkedIn Posts or Short Videos

Our expert product team has selected these resources to help you save thousands of dollars on unnecessary subscriptions, avoid wasting time reading instead of taking action, and save years following the wrong advice. Our mission is to introduce you to brilliant product thinkers, whether they’re renowned experts or emerging voices.

---

Top Long-Form Content

1. How to Turn Your AI into a World-Class Designer

   1. What we liked: An ex-Apple design lead lays out an end-to-end process for pulling real creative work out of a model instead of generic slop.

   2. Choice quote: "Most people only see 1% of AI's creative potential."

2. The AI Implementation Gap: Why Most AI Projects Never Reach Production

   1. What we liked: Calls out exactly why so many AI pilots aren’t ready for production and how to avoid this trap.

   2. Choice quote: "The uncomfortable pattern underneath all of this is that pilots and production systems are optimized for entirely different things, and companies keep mistaking success in one for progress toward the other.."

3. Faster at What Exactly?

   1. What we liked: Challenges the assumption that faster with AI is better.

   2. Choice quote: "What if the part I skipped to save time was the part where I would have noticed that I was asking the wrong question?”

Top Tweets

1. Tokens per Megawatt: A New Yardstick for Inference Efficiency

   1. What we liked: An relatively accurate yet humorous calculation showing how human speech compares to AI speech in terms of energy impact.

   2. Choice quote: It’s a formula. :)

2. The Morning ChatGPT and Claude Both Went Dark

   1. What we liked: Two labs, same morning outage, and the observation that almost nothing built on these models has a fallback when they stop.

   2. Choice quote: "Everyone's pricing AI like it's already infrastructure."

3. Computer-Use AI Is About to Blow Up Compute Demand

   1. What we liked: A clear macro thesis from a credible markets voice, relevant to anyone forecasting AI cost curves into their roadmap.

   2. Choice quote: "AI compute demand is going to go exponential again."

Top LinkedIn Posts & Short Videos

1. Humans Blindly Approved 97% of Claude Code Prompts

   1. What we liked: The sharpest governance insight in the batch — human-in-the-loop quietly collapses once prompt frequency gets high enough. Essential for anyone designing agent approval flows.

   2. Choice quote: "Human reviewers were already approving 97% of permission prompts without reading them."

2. They Built Instead of Bought. Here Comes the Bill.

   1. What we liked: The build-vs-buy decision through a procurement lens nobody else in the batch brings. Directly useful for founders and product leaders weighing agentic builds.

   2. Choice quote: "For builds, that TCO hides in cloud invoices, headcount for maintenance, and opportunity cost. For buys, it hides in implementation fees (20-30% of sticker price), renewals, and consumption overages.."

3. Governance Is the Moat

   1. What we liked: Enjoyed the framing that telling AI when to stop within a task is a major new product moat.

   2. Choice quote: "So we stopped asking what the agent can do, and started asking what it is allowed to do."

4. Stop Calling It "AI Agent" If It Can't…

   1. What we liked: An AI PM’s crisp line between actually-agentic and “uses an LLM somewhere” with some healthy pushback on marketing. That said, this article goes too far and misses that a lot of advanced chatbot systems are very much agentic.

   2. Choice quote: "Most stuff that marketers call 'agentic' is not.
      They’re a workflow with an LLM in the middle."

---

This article is brought to you by Productboard

Trusted by the world’s best product teams, Productboard turns customer needs into high-impact strategy in one end-to-end platform. With Productboard Spark, you can go from raw idea to polished product brief in seconds using AI that already understands your business.

See how Productboard Spark can help your product teams ship faster →

---

In case you missed them — Adam’s recent LinkedIn Content

1. The company selling an “AI Business Development Representative” is hiring a human BDR.

2. "Don't worry, a human checks the AI before anything happens."

3. There are two types of salespeople no CEO wants to hire.